[Bug 58396] [i945GM regression] in 3.7.0 kernel on macbook1,1 (kernel crash)

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=58396

--- Comment #25 from Daniel Vetter <daniel at ffwll.ch> ---
> --- Comment #24 from Peter Ujfalusi <peter.ujfalusi at gmail.com> ---
> (In reply to comment #22)
>> Please test the attached patch. I still need to decide whether this is the
>> "right" fix, but I think this should at least prevent your oops.
>
> With this patch KMS also works. If you decide that this is the right fix can
> you please make sure it hits linux-stable (3.7) soon.

I'll submit it with the bug link updated asap for 3.8, if it passes
review and all, it should land in a stable backport in 1-2 weeks. But
holidays might delay that a bit.

> Out of curiosity: do you have short explanation on why this patch actually
> helps/works?

Your BIOS enables VGA on the same pipe 1 as it enables LVDS. When
setting up the initial kms config we go through all pipes: First we
disable pipe 0, since it's unused, then we'll enable the desired
config for LVDS on pipe 1.

The problem now is that for historical reasons the new i915 modeset
code disables any outputs when they're disconnected for _any_ modeset,
i.e. when we ensure that pipe 0 is off, it also wants to disable VGA
out on pipe 0. Imo this is braindead semantics, but it's bug-for-bug
compatible with the old modeset implementation from 3.6. The problem
now is that the new modeset always enables/disables the entire pipe,
so to disable VGA on pipe it needs to enable/disable LVDS on pipe 1,
too. There we fall over, since at we don't (yet) read out the mode set
up by the BIOS, hence the NULL deref oops.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20121218/d2583a07/attachment.html>